NAMD hybrid configuration: Atoms moving too fast error

From: Puneet Singh (singh.punit1990_at_gmail.com)
Date: Mon Jul 27 2015 - 06:41:38 CDT

Dear Sir/Madam,
I scanned through the archive list ,but i was not able to find the exact
resolution to this issue. I am trying to build NAMD for host only (without
mic/gpgpu support) hybrid configuration (OpenMP/threaded + MPI) .But i am
having trouble running apoa1.namd example with namd2 smp binary. Brief
build procedure & errors encountered are as :-

1. I am compiling the charm (with smp) and then compiling NAMD with intel
mpi compilers.

NAMD smp build script is attached :* namd_build_smp.sh* .

The script generated namd2 binary successfully, but when i try to run the
namd2 + apoa1.namd example i get:
ERROR: Atom 27216 velocity is -10752.9 1207.92 -5988.05 (limit is 12000,
atom 268 of 667 on patch 64 pe 18)
ERROR: Atoms moving too fast; simulation has become unstable (1 atoms on
patch 64 pe 18).
(*attached*: *namd_with_smpError.log*)

2.
 a) Then i again tried to build the non-smp charm (attached: namd_build.sh),
then i was able to run the apoa example using:-

*mpirun -np 23 ./namd2 ./apoa1/apoa1.namd*
(*attached*: *namd_with_nonsmp.log*)
This implies that the input file is okay , the troubles are with my smp
build procedure (i guess).

 b) But with non-smp version if i try to run it as multithreaded/smp using:-
*mpirun -np 1 ./namd2 +ppn 23 ./apoa1/apoa.namd*
(*attached:* *namd_with_nonsmpError.log*)
------- Partition 0 Processor 0 Exiting: Called CmiAbort ------
Reason: +ppn cannot be used in non SMP version!
which is obvious.

I want to know that:-
Why the smp configuration fails while non-smp configuration succeeds ,
weather my build procedure or the input file is at fault? i.e.

   - What improvements are needed in the build procedure (namd_build_smp.sh)
   so that i cam run the apoa1 example.
   - or, What modifications are needed in input file apoa1.namd example
   to make it work on hybrid configuration(smp) without getting Atom
   moving fast error ?

Any help/Hint/references will be very fruitful.

Eagerly awaiting your reply,

Thanks & Regards,
Puneet Singh


This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:21:15 CST